Add Guidelines

Guidelines for syntax inside prompt

I think it is somewhat important to write a blog-page or little video on front-page of dashboard or when creating an account to explain what should be done and what shouldn't be done.

Some other assistant like ChatGPT or bard uses markdown inside the prompt to display content such as

• Lists
• italic , bold, not you, italic-bold
• Links

• quote

 x = "code with color"

• Footnot [^1],
• And many more...

What behavior do we want Open-Assistant to have ?

I already started to add markdown in my answers as the assistant, should I ? Will it help training ?

What about links to the web (if the prompter asks for specific information that I feel the need to add source) ? Can I add links ?
For example, a prompter asks specifically what are the performance of a rtx-4090 versus a rx-7800 to run Stable Diffusion, in the answer, I added benchmarks from different source such as this one.

In this roadmap, you mention third party and retrieval of information, how will this be implemented ? Wouldn't it be preferable to already train answers/prompt using same syntax? If it's markdown, then maybe add some JS textbox with preview like what GitHub does:

This guideline page could also mention behavior for specific topics such as general medicine, or urgency. Should we by default reject all medical advices ? What about basic information such as telling to someone doing panic attack to breath, or what to do to someone having seizure or stroke ?

[^1]: This footnot is kinda cool, but can make copy/past hard as there is [^1] lot [^1] of [^1] reference [^1] like [^1] in wikipedia [^1][^1][^1]

The issue names no repository files or tests. Review the proposed Markdown examples, the linked roadmap, and the dashboard or account-creation locations mentioned; done means the project has agreed guidance covering prompt syntax, links, rendering, and the medical-advice cases described.